Statistical pre-processing method for peripheral quantitative computed tomography images

Abstract

This study aimed to find a processing method that would reduce the noise level and enhance the image quality for structural bone analysis in peripheral quantitative computed tomography (pQCT) images. We proposed method based on down-sampling of histogram of gray scale intensities and following correction of subsequent inaccuracies. It employs wavelet transform and a Markov random field model. For comparison, two well known techniques for filtering of images (median filtering and filtering based on wavelet transform with using soft-thresholding) were evaluated. The performance of the used methods was tested on pQCT scan of artificial phantoms, the real pQCT scan of distal tibia as well as on numerical model of pQCT scan. As to the preservation of coarse structural information of pQCT images, it seems that the new preprocessing method based on statistical approach performed reasonably well and appears to be a promising method for enhancing the analysis of pQCT images. © 2010 International Federation for Medical and Biological Engineering.